Samsung SSD 850 Pro (128GB, 256GB & 1TB) Review: Enter the 3D Era

Over the last three years, Samsung has become one of the most dominant players in the SSD industry. Samsung’s strategy has been tight vertical integration ever since the beginning, which gives Samsung the ability to be in the forefront of new technologies. That is a massive advantage because ultimately all the parts need to be designed and optimized to work properly together. The first fruit of Samsung’s vertical integration was the SSD 840, which was the first mass produced SSD to utilize TLC NAND and gave Samsung a substantial cost advantage. Even today, the SSD 840 and its successor, the 840 EVO, are still the only TLC NAND based SSDs shipping in high volume. Now, two years later, Samsung is doing it again with the introduction of the SSD 850 Pro, the world’s first consumer SSD with 3D NAND.

For years it has been known that the scalability of traditional NAND is coming to an end. Every die shrink has been more difficult than the previous as the endurance and performance have decreased with every node, making it less and less efficient to scale the size down. Scaling below 20nm was seemed as a major obstacle but the industry was able to cross that with some clever innovations in the NAND design. However, the magic hat is now running out of tricks and a more signficant change to the NAND design is required to keep scaling the cost.

The present solution to the scalability problem is 3D NAND, or V-NAND as Samsung calls it. Traditionally NAND and other semiconductors are scaled horizontally along the X and Y axes but due to the laws of physics, there is a limit of how small the transistors can be made. To solve the problem, 3D NAND introduces a Z-axis i.e. a vertical dimension. Instead of cramming transistors horizontally closer and closer to each other, 3D NAND stacks layers of transistors on top of each other. I will be going through the structure and characteristics of 3D NAND in detail over the next few pages.

By stacking transistors (i.e. cells when speaking about NAND) vertically, Samsung is able to relax the process node back to a much more convenient 40nm. When there are 32 cells on top of each other, it is obvious that there is no need for a 10nm-class node because the stacking increases the density, allowing production costs to scale lower. As we have seen with the history of NAND die shrinks, a higher process node provides more endurance and higher performance, which is what the 850 Pro and V-NAND is all about.

Fundamentally the only change in the 850 Pro is the switch to V-NAND. The interface is still SATA 6Gbps and the controller is the same triple-core MEX from the 840 EVO, although I am still waiting to hear back from Samsung whether the clock speed is the same 400MHz. The firmware, on the other hand, has gone through a massive overhaul to adopt the characteristics of V-NAND. With shorter read, program and erase latencies and higher endurance, the firmware needs to be properly optimized or otherwise the full benefits of V-NAND cannot be utilized.

I bet many of you would have liked to see the 850 Pro move to the PCIe interface but I understand Samsung’s decision to hold off with PCIe for a little while longer. The market for aftermarket PCIe SSDs is still relatively small as the PC industry is figuring out how to adopt the new interface, so for the time being Samsung is fine with watching from the side. The XP941 is and will continue to be available to the PC OEMs but for now Samsung will be keeping it that way. From what I have heard, Samsung could bring the XP941 to the retail market rather quickly if needed but Samsung has always been more interested in the high volume mainstream market instead of playing in the niches.

The performance figures in the table above give us the first glimpse of what V-NAND is capable of. Typically modern 128GB SSDs are only good for about 300MB/s but the 850 Pro is very close to saturating the SATA 6Gbps bus even at the smallest capacity. This is due to the much lower program times of V-NAND because write performance has been bound by NAND performance for quite some time now.

The other major improvement from V-NAND is the endurance. All capacities, including the smallest 128GB, are rated at 150TB, which is noticeably higher than what any other consumer-grade SSD offers. Moreover, Samsung told me that the endurance figure is mainly meant to separate the 850 Pro from the enterprise drives to guide enterprise clients to the more appropriate (and expensive) drives as the 850 Pro does not have power loss protection or end-to-end data protection for example. However, I was told that the warranty is not automatically denied if 150TB is reached under a client workload. In fact, Samsung said that they have a 128GB 850 Pro in their internal testing with over eight petabytes (that is 8,000TB) of writes and the drive still keeps going, so I tip my hat to the person who is able to wear out an 850 Pro in a client environment during my lifetime.

Another interesting aspect of V-NAND is its odd capacity per die. Traditionally NAND capacies have come in powers of two, such as 64Gbit and 128Gbit, but with V-NAND Samsung is putting an end to that trend. The second generation 32-layer V-NAND comes in at 86Gbit or 10.75GB if you prefer the gigabyte form. I will be covering the reason behind that in more detail when we look at V-NAND more closely in the next few pages but as far as I know there has never been a strict rule as to why the capacities have scaled in powers of two. I believe it is just a relic from the old days that has stayed in the memory industry because deep down binary is based on powers of two but the abnormal die capacity should have no effect on the operation of the NAND or the SSD as long as everything is optimized for it.

Due to the odd die capacity, the die configurations are also quite unusual. I found two different capacity packages inside my review samples and with Samsung’s NAND part decoder I was able to figure out the die configurations for each capacity. Unfortunately, Samsung did not send us the 512GB model and I could not get the 128GB model open as Samsung uses pentalobe Torx screws and I managed to wear out the screw while trying to open it with an inappropriate screw driver (it worked for the other models, though), so thus there are question marks at those capacities in the table. However, this should not impact the raw NAND capacities as long as all capacities follow the same 7.6% over-provisioning trend but the package configurations may be different. I will provide an update once I receive a confirmation from Samsung regarding the exact configurations for each capacity.

The 850 Pro also switches to smaller PCB designs. The PCB in the 1TB model populates around two thirds of the area of the chassis, while the 256GB PCB comes in at even smaller size. The reason for the different PCB sizes is the amount of NAND packages as the 256GB only has four, whereas to achieve the capacity of 1TB eight NAND packages are required.

New Samsung SSD 840 EVO Read Performance Fix Coming Later This Month

The Samsung SSD 840 EVO read performance bug has been on the table for over six months now. Initially Samsung acknowledged the issue fairly quickly and provided a fix only a month after the news hit the mainstream tech media, but reports of read performance degradation surfaced again a few weeks after the fix had been released, making it clear that the first fix didn’t solve the issue for all users. Two months ago Samsung announced that a new fix is in the works and last week Samsung sent us the new firmware along with Magician 4.6 for testing, which will be available to the public later this month.

I covered the reason behind the issue in one of our earlier articles, but in short the read performance degradation is a result of cell charge decay over time that caused extensive read-retry cycles to retrieve the correct data. The new firmware fixes this by periodically refreshing (i.e. rewriting) old data, which recovers the cell charge back to its original state and ensures that no read-retry or ECC that would degrade the performance is needed. Samsung says that the refresh operation does not impact user performance, suggesting that it’s a relatively low priority process that is run when the drive is idling.

The new Magician 4.6 also includes an Advanced Performance Optimization feature, which is similar to the performance restoration tool that Samsung released earlier. Basically, it’s a command that tells the SSD to rewrite all of its internal data, which resets all cell charges and hence recovers performance. It’s merely a supplementary tool as the firmware upgrade itself should be enough to restore performance, but in case the performance isn’t fully restored after the firmware upgrade (and some idle time to let the drive refresh the cells), the tool can be used to force a cell charge refresh.

I haven’t run any tests of my own because I don’t have any 840 EVOs deployed in my systems (I only have one 2.5″ EVO anyway), but Allyn Malventano from PC Perspective managed to run some tests on a degraded drive to show the impact of the new firmware.

Allyn’s tests indicate that the new firmware seems to mostly fix the issue even without running the optimization tool. Note that Allyn didn’t give the drive any idle time after the firmware update, so the update appears the be very effective and with idle time the performance would likely have restored on its own.

Obviously, the big question is whether the performance will stay high because there was never a problem with freshly written data. We won’t know that for sure until a couple of months later, but given the way the new firmware handles old data it does sound more promising because no data should get old enough to be slow to read.

Some of you are likely skeptical about the effect on endurance since rewriting the data will consume P/E cycles, but I find this to be a non-issue. We know that Samsung’s 19nm TLC NAND is rated at 1,000 P/E cycles, so if the drive was to refresh all cells once a week, even that would only consume 52 cycles in a year. In five years time the total would be 260 cycles, which leaves you with 740 cycles for user data writes (for the record, that’s 52GB of NAND writes per day for five years with the 120GB 840 EVO).

All in all, I hope this fix will finally put an end to the performance degradation. The issue has been bugging many users for months and it’s critical that the users get what they initially paid for. On one hand I’m confident enough to say that this fix is permanent given the way it works, but on the other hand I don’t want to be too optimistic this time around because the first fix didn’t turn out so great. Either way, I think this fix is the last chance for Samsung to provide a permanent solution because they already failed to do so once and it would no longer be fair to ask the customers to wait months for a fix that might or might not fix the issue. For now the only thing we can do is wait for user reports and hope for the best, but at least in theory the new firmware should be a permanent fix.

Samsung SM951-NVMe (256GB) PCIe SSD Review

PCIe and especially NVMe SSDs are without a doubt the hot topic in the SSD industry at the moment. There are still only a handful of drives on the retail market, but as we saw at Computex a few weeks ago, everyone is working closely on their PCIe designs and we should see more entries to the market later this year with a big wave of PCIe SSDs arriving in the first half of 2016.

Samsung has always been an early adopter in the SSD space. The company was the first one on the market with a PCIe 2.0 x4 M.2 SSD the (XP941) back in late 2013, and before that it was the first one to adopt TLC NAND in 2012. Earlier this year Samsung’s second generation client PCIe drive, the SM951, made an appearance in a Lenovo laptop, but to everyone’s surprise the drive wasn’t NVMe compatible like Samsung had announced earlier. After discussing with Samsung, the company said they it has an NVMe client drive in development, but it declined to provide any reasoning as to why the SM951 still used the AHCI driver stack.

To our surprise, Ganesh found an NVMe-enabled Samsung M.2 SSD inside Intel’s Broadwell-U NUC a while back. This was rather confusing at first because Samsung had specifically told us that the SM951 doesn’t support NVMe, but after a closer look and a series of emails with Samsung the drive turned out to be an NVMe version of the SM951, or SM951-NVMe as Samsung calls it.

Distinguishing the AHCI and NVMe version from each other isn’t very simple as the difference lies in a single character in the model number. The AHCI version carries the code MZ-HPVxxx0 (where xxx is the capacity in gigabytes), whereas the NVMe version is called MZ-VPVxxx0. Since both versions of the SM951 are technically OEM-only, the close naming isn’t really an issue, but if you are shopping for the SM951 I recommend that you take a close look at the part number before pulling the trigger to ensure that you get the version you are looking for.

On the hardware side the AHCI and NVMe versions of the SM951 are a match. Both utilize Samsung’s S4LN058A01-8030 controller dubbed as UBX, which is a PCIe 3.0 x4 controller that apparently supports both AHCI and NVMe driver stacks. That isn’t surprising, though, because nearly all client-grade NVMe controllers I know are capable of supporting both — it’s just a matter of developing two separate firmware builds. The firmware development is likely the reason why the SM951-AHCI was the first one to market because Samsung already had the basic AHCI firmware from its XP941 and SATA drives, whereas the SM951-NVMe needed more development from scratch given how different and more efficient the NVMe command set is.

Similar to the AHCI version, the SM951-NVMe only comes in capacities of up to 512GB. The reason lies in NAND because the SM951 utilizes 64Gbit dies, and with only four NAND package placements on the M.2 2280 PCB the maximum capacity with 16 dies per package works out to be 512GB (8GB x 16 x 4). It seems that Samsung doesn’t have a high volume 128Gbit MLC die at this point, although we will likely see one with third generation V-NAND later this year. The first generation V-NAND die is 128Gbit, but since it only has 24 layers it’s not cost efficient for a client drive, especially not for an OEM-specific one given how cost sensitive PC OEMs are.

TechInsights found out that the NAND in the SM951 (both AHCI and NVMe) is actually 16nm. While I was aware of the change in generation character of the part number, I believed that it would just be a second generation of Samsung’s 19nm die because to me it didn’t make any sense that Samsung would build a 16nm 64Gbit die. I’m working on an article comparing all the modern 15-16nm NAND processes, so stay tuned for more in-depth analysis of Samsung’s 16nm node.

Boot Support
One of the major questions with every PCIe SSD is whether it is bootable. Back when the XP941 became available the situation was rather messy because motherboard OEMs had not prepared for PCIe SSDs yet, which require BIOS/UEFI support from their side in order to show up in the boot menu. Fortunately, most OEMs fixed this for 9-series motherboards and now most models have a BIOS update available with proper support for PCIe and NVMe SSDs.

In short, the SM951 NVMe is bootable in my ASUS Z97 Deluxe with the latest 2401 BIOS. I don’t have any other 9-series motherboards at hand, but I suspect that any motherboard with advertised NVMe support and appropriate BIOS will boot from the SM951 NVMe.

For tower Mac Pro users the story isn’t as pleasant, though. I put the SM951 NVMe inside my 2009 Mac Pro, but OS X wouldn’t even recognize the drive. Despite the fact that the custom Apple SSD inside the MacBook is NVMe based, I suspect that the current version of OS X doesn’t carry a general NVMe driver, and even if it did the Mac Pro and its chipset might simply be too old to support NVMe, which honestly isn’t surprising for a +5-year-old system. In any case, Mac Pro users can still buy and boot from the AHCI version of SM951, but I wouldn’t hold my breath for any NVMe support in the future.

The SM951-NVMe is an OEM part, meaning that availability is very restricted. The drive is listed by a handful of online retailers, but none of them seem to have it in stock yet. RamCity is expecting stock in mid-July, but told us that even that is uncertain because its distributors are still saying that the NVMe version is in sampling stage with no schedule for high volume availability. We got our 256GB sample directly from Samsung, hence the early access, as it seems that there is no way to buy the SM951-NVMe at this point. I will provide an update when I hear more about the availability.

Intel Announces SSD DC P3608 Series

Intel is introducing a new family of enterprise PCIe SSDs with the aim of outperforming their existing DC P3600 series and even beating the DC P3700 series in many metrics. To do this, they’ve essentially put two P3600 SSDs on to one expansion card and widened the interface to 8 lanes of PCIe 3.0. While this does come across as a bit of a quick and dirty solution, it is a very straightforward way for Intel to deliver higher performance, albeit at the cost of sharply increased power consumption.

The SSD DC P3608 appears to the system as two individual NVMe drives behind a PLX PCIe switch chip. This means that extracting full performance from this card will require software RAID-0 or some similar software load-balancing solution. A new version of Intel’s Rapid Storage Toolkit for Enterprise (RSTe) drivers will be providing this capability. The overhead of the PCIe switch and managing two independent controllers means that the P3608 cannot attain an oughtright doubling of the P3600’s performance.

The inclusion of two SSD controllers and a PCIe switch chip also drives idle power consumption up to 11.5W and makes a 2.5″ form factor impossible, so the P3608 series will only be available as a half-height half-length PCIe expansion card. Intel’s not too worried about the form factor constraint, because they’re now able to make full use of the 8-lane PCIe slots that are the most common in the sort of servers these drives are typically used in.

The SSD DC P3608 is available in three capacities, with the smallest 1.6TB configuration having more overprovisioning to boost random write speeds. Active power consumption varies with capacity, but all models support a power governor setting to limit power consumption to 35W or 25W instead of the worst-case 40W. Intel has provided us with a 1.6TB SSD DC P3608, so a full review is on its way.

Micron 3D NAND Status Update

Update: We’ve got some more information and diagrams from Micron’s Winter Analyst Conference earlier today.

After samples of their upcoming 3D NAND were sighted in the wild at CES, Micron has taken the time to provide some details about the flash memory and their plans for it. A lot of this is a recap of information we’ve previously covered, but we’ve got some new details and a better idea of the roadmap for the future.

The entire flash memory industry has shifted focus to the devlopment of 3D NAND flash memory as the replacement for planar NAND flash memory. Samsung took an aggressive approach and has enjoyed some great success with their V-NAND branded 3D NAND, but it hasn’t been an entirely trouble-free transition. Micron has been more conservative both in technology and timing, but they plan on having a strong competitor on the market later this year.

Micron’s first generation 3D NAND takes the form of a 256Gb MLC die and a 384Gb TLC die (compare with their 128Gb 16nm MLC and TLC). At a high level, the die will be partitioned into four separate planes, compared to two planes for most competing NAND. A 480GB drive using the four-plane 256Gb dies will have access to approximately the same amount of parallelism as a 480GB drive using two-plane 128Gb dies, so this capacity jump won’t bring the performance drops that have tarnished some NAND process shrinks.

The key development that allows Micron to produce a four-plane die without inflating die size and cost relative to the two-plane competition is that they’ve layered much of the required additional circuitry under the 3D flash array, instead of sitting alongside. Micron says that their “CMOS Under the Array” design puts more than 75% of the logic (things like address decoding and page buffers) under the flash memory. It doesn’t make the additional segmentation of the four-plane design entirely free, but it allows it to be a very cost effective performance optimization. This is still planar CMOS logic, not any kind of 3D or stacked logic; it’s just got some metal interconnect layers and the flash array piled on top.

On a smaller scale, the 3D NAND will have a page size of 16kB and erase block sizes of 16MB for the MLC and 24MB for the TLC. Because CPUs and filesystems are still mostly dealing with 4kB chunks, Micron has included a partial page read capability that allows for a 4kB read to be done a bit faster and with about half the power of a full 16kB page read. This helps offset some of the penalty the larger page size can have on random 4kB read performance. The large erase block sizes won’t have much of a direct impact on performance and are a necessary efficiency measure: erasing requires charge pumps to produce higher voltages than reads or writes use, and it’s a slower and more power-hungry operation. If you’re going to fire up that extra circuitry and block access to the entire plane for 1ms or more, you might as well erase a usefully large amount of flash.

For the architecture of the individual memory cells, we have nothing new to report. Intel and Micron are alone in their decision to stick with floating-gate flash technology instead of transitioning to charge-trap flash. We’ve previously explained how the technologies differ and what kinds of advantages the manufacturers want to reap from the change. The cost is that the design process involves different tradeoffs that are not as thoroughly explored and understood as the dynamics of floating-gate flash, and for now Micron is sticking with what they know. Micron’s 3D NAND might not have the best write endurance, but they’re expecting to have an advantage in data retention time for healthy flash. They aren’t providing exact numbers, but they’re estimating that drives relying on simpler BCH ECC can get effective program-erase cycle lifetimes in the thousands and drives with LDPC will have effective cycle counts of tens of thousands. Once the process has matured it should exceed their 20nm planar NAND’s write endurance.

The first 3D NAND Micron has ready for the market will produced to the endurance standards for client drives, with enterprise-grade 3D NAND following later. The MLC is currently a few weeks ahead of the TLC in the qualification process, but given the state of the client SSD market the TLC will be the more popular product and will overtake the MLC in volume produced within a few months of 3D NAND drives hitting the market. Overall their 3D NAND will comprise a majority of their flash output on a per-GB basis by the second half of 2016. Micron is sampling drives with 3D NAND to partners this month and is planning for general availability in June. Other drive vendors using Micron’s NAND will be on similar release schedules.

Micron hasn’t announced any specific drive models, but they’ve given a general roadmap that is unsurprising. Consumer and client products will come during the middle of the year, with the capacity and cost improvements allowing for things like 2TB 2.5″ drives and 1TB single-sided M.2 drives. Toward the end of 2016 and into 2017 we’ll see enterprise products such as very high capacity (8TB+) drives and updates in the existing product segments for SAS and PCIe drives.

Looking further to the future, Micron gave a presentation last week at the IEEE International Solid-State Circuits Conference entitled “A 768Gb 3b/cell 3D-Floating-Gate NAND Flash Memory”. This was more about bragging about their R&D in an academic context than announcing any concrete future product plans, but it does represent the most likely successor to their first-generation 3D NAND. The chip in question provides a whopping 768Gb (96GB) capacity when operated as TLC, and 512Gb (64GB) as MLC. The die size is about the same as their 32-layer 384Gb TLC, the areal bit density is almost doubled, and most of the other details are the same—implying that the layer count has probably increased, though Micron isn’t saying how many layers it uses. If Micron has plans to switch to charge-trap flash they’re keeping it under wraps for now, and any such transition isn’t imminent. The second-generation 3D NAND will start production in their Singapore fab this summer, and volume will be ramping up around the end of 2016 (during the second quarter of their fiscal year 2017). Micron predicts their second-generation 3D NAND will be at least 30% cheaper per Gb than the first generation, which they report to be at least 25% cheaper than their 16nm planar NAND.

Samsung Shows Off SM961 and PM961 SSDs: OEM Drives Get a Boost

At Samsung’s annual SSD Forum Japan, the company has demonstrated two previously-unannounced high-performance client SSDs. The new SM961 and the PM961 drives are based on the company’s Polaris controller as well as V-NAND flash memory. Samsung promises that the SSDs will increase sequential read speeds to 3000 – 3200 MB/s and will also significantly boost random read and write performance. The drives are projected to ship inside PCs in the second half of the year, while it’s anyone’s guess if and when these will filter into retail (ala the 950 series).

The Journey Continues
Samsung is the world’s largest producer of NAND flash memory and SSDs. Moreover, it is also one of the companies, which raise the bar of SSD performance and feature-set. In the recent years, Samsung has been an instrumental driving force in popularization of PCIe and NVMe SSDs. Samsung’s XP941 was one of the first high-end PCIe M.2 SSDs aimed at OEMs and it greatly helped to promote the form-factor among PC makers in 2014. The SM951 SSDs significantly increased performance of flash storage sub-systems compared to its predecessor and introduced NVMe to numerous OEMs a year later. In fall 2015, the Samsung 950 Pro finally brought NVMe and performance of the SM951 to retail market.

When introduced, the XP941 and the SM951 were for their respective times Samsung’s top-of-the-range SSDs designed for OEMs and their flagship PCs. Last year, the company decided to alter its strategy regarding high-performance SSDs aimed at PC makers. Instead of offering just one lineup of fast PCIe drives, Samsung introduced its second family of advanced SSDs with lower price and TLC NAND — the PM951. While the latter drives were still rather fast, the price/performance trade-off meant that they were considerably behind the SM951 in terms of sequential read and write speeds. This year, the company plans to change its approach to OEM SSDs once again. Samsung will offer two product families with slightly different characteristics: the SM961 and the PM961. The new lineups will further raise performance bar of Samsung’s SSDs, but the PM961 is expected to make the new speed levels accessible to a broader audience. The tactics should help Samsung to increase its share of the premium SSD market.

Samsung SM961 and PM961: First to Use Polaris Controller
The new OEM SSD lineups from Samsung will be the company’s first standard high-end SSDs for PC makers, which use 3D/V-NAND memory. The new drives rely on the all-new Polaris platform and will come in M.2-2280 form-factor with PCIe 3.0 x4 interface while utilizing the NVMe protocol. Right now, the company is not revealing too many details about its new SSDs. For example, it is unknown whether they rely on Samsung’s third-gen 48-layer V-NAND or its second-gen 32-layer V-NAND. While using newer memory for new SSDs is logical, so far Samsung has not confirmed anything.

The Samsung SM961 will be Samsung’s new top-of-the-range M.2 SSD line for OEMs, which will be offered in 128 GB, 256 GB, 512 GB and 1 TB configurations (by contrast, the SM951 family did not include a 1 TB option). The drive will be based on Samsung’s MLC V-NAND as well as the company’s Polaris controller. Samsung is specing the SM961 at up to 3200 MB/s for sequential reads and up to 1800 MB/s for sequential writes, but does not specify which models will boast with such numbers. The new SSDs can perform up to 450K random read IOPS as well as up to 400K random write IOPS, which looks more like performance of server-grade SSDs.

The Samsung PM961 will be based on the company’s TLC V-NAND flash and the Polaris controller. The PM961 lineup will consist of four models: with 128 GB, 256 GB, 512 GB and 1 TB capacities (the PM951 also did not include a 1 TB model). The PM961 SSD supports sequential read speeds of up to 3000 MB/s as well as sequential write speeds of up to 1150 MB/s. Random read/write performance of the PM961 is up to 360K/280K of read/write IOPS, but Samsung does not specify exact models that offer such performance.

The PM961 should be more affordable than the SM961, but even with the use of TLC, according to Samsung’s specifications it should be only slightly slower than the flagship model when it comes to sequential read speeds (6% difference is negligible). Sequential write speeds should be lower compared to those offered by the SM951 and the 950 Pro, but should still be considerably higher than sequential write speed of its direct predecessor, the PM951. Moreover, the new PM961 should also be considerably faster in random read and write operations versus its ancestors, according to specifications released by Samsung.

It remains to be seen how the PM961 drive will behave in real-world applications versus enthusiast-class PCIe SSDs from other manufacturers. But its 3000 MB/s sequential read performance as well as very high random reads and writes should probably play positive roles here.

Samsung has not formally introduced the SM961 and the PM961 SSDs, but PC Watch reports that the company expects computer makers to use the new drives in the second half of the year. Moreover, since the SSDs demonstrated at the event already feature actual labels with all the certification stamps and even serial numbers, it looks like development of the storage devices is very close to its completion.

Initially, Samsung will use its Polaris platform only for the drives intended for OEMs, but it is plausible to expect the company to utilize the same controllers for its retail SSDs eventually. As is usually the case, Samsung isn’t commenting on whether we’ll see retail Polaris drives in the future, though if 950 Pro is anything to go by, we wouldn’t be surprised. Nonetheless, keeping in mind how easy is to get both SM951 and PM951 SSDs from stores like Amazon and RamCity, it should not be a problem to obtain these new OEM SSDs from online stores as well.

Looking Ahead: PCIe 3.0 x4 May Become a New Performance Limiting Factor for SSDs
Looking at Samsung’s specifications, it’s interesting to note that the 3200 MB/s sequential read speed of the Samsung SM961 is very close to actual maximum bandwidth of the PCIe 3.0 x4 bus. A PCIe 3.0 connection supports transfer rates of 8 GT/s (GigaTransfers per second) per lane, for a total of 3.94 GB/sec for an x4 connection after factoring in overhead. As it appears, the bandwidth offered by four lanes of PCIe 3.0 – the maximum bus width supported by the M.2 standard – is close to being saturated by high-end client SSDs.

If this ends up being the case, then the SM961 may be Samsung’s fastest sequential transfer SSD for client PC applications for quite a while (i.e., until PCIe 4.0 arrives). Consequently, if the company plans to introduce its successors with the same PCIe 3.0 interface, it will have to improve other domains of SSD performance (e.g., performance in mixed workloads, endurance, power efficiency, etc.). Keeping in mind that Samsung’s retail 950 Pro SSDs were a little faster than the company’s SM951 despite using the same controller, it will be interesting to see how Samsung might differentiate its hypothetical enthusiast-oriented Polaris and MLC V-NAND-based SSD from its OEM offering (if such SSD is in the company’s plans at all, of course).

With that said, while certain Samsung’s SSDs may eventually hit a sequential transfer performance barrier in the form of PCIe 3.0 x4 bus, the bandwidth provided by this interface will still be enough for the vast majority of SSDs from different manufacturers going forward. Moreover, since real-world performance does not entirely depend on just sequential read or write speeds, there are a plenty of ways to improve performance of actual drives in real-world applications.

Evaluating the Toshiba OCZ RD400 M.2 NVMe SSD on a Skylake NUC

The Skylake NUCs have faced plenty of issues since their introduction late last year. However, stability has improved with the latest BIOS updates. One of the issues with the early BIOS versions was the M.2 SSD slot being effectively limited to PCIe 2.0 rates despite the chipset supporting PCIe 3.0. The latest BIOS version resolves this issue. We had carried a detailed evaluation of different M.2 SSDs for usage in the Skylake NUC. At that time, the Samsung SSD 950 PRO was the only option for providing maximum possible theoretical performance – a NVMe SSD with a PCIe 3.0 x4 link. Since then, Toshiba has also entered the fray with the Toshiba OCZ RD400. The RD400 has already been put through our rigorous SSD evaluation suite. In this review, we take a look at how the 512GB version fares in the NUC6i5SYK.

Introduction to the Toshiba OCZ RD400
The Toshiba OCZ RD400 is a PCIe 3.0 x4 NVMe SSD in the M.2 form factor. It can be purchased either in standalone M.2 2280 form factor, or with an add-in card. We have already discussed the RD400 hardware in detail. The important takeaway is that the controller is likely to be the Marvell 88SS1093, though we do not have direct confirmation of this aspect.

The 512GB version we evaluated is a single-sided card with two flash packages, a single DRAM package and the controller on board. These packages are covered by a sticker with the OCZ logo. In the Skylake NUC, the thermal strip attached to the lid is able to make contact with this sticker to aid in cooling down the components.

OCZ’s site provides a special driver that improves performance compared to the standard Microsoft NVMe driver that ships with Windows 10. The SSD Utility software also gives an overview of the current state of the drive (including the currently active driver).

There are plenty of SSDs compatible with the Intel NUC6i5SYK, but, what does one choose for the best experience? There are decisions to make – PCIe 3.0 vs. PCIe 2.0, NVMe vs. AHCI, PCIe vs. SATA and so on. In this review, we are going to compare the Toshiba OCZ RD400 against four other M.2 SSD options:

Intel Launches 3D NAND SSDs For Client And Enterprise

Today Intel is announcing a variety of new SSDs with their 3D NAND flash memory. The new models use a mix of 3D MLC and 3D TLC, some SATA and some PCIe, and variously target the consumer, business, embedded and data center markets. While we are still awaiting details on the timing of these product releases, it is clear that Intel is eager to put planar flash behind them. The drive for this is especially strong as the models being replaced are all either based on Intel’s relatively expensive 20nm flash or on 16nm flash that Intel had to buy on the open market due to their decision to not participate in the 16nm node at IMFT.

First up, we have a M.2 PCIe SSD branded three different ways for three different markets. In the consumer market we have the SSD 600p series, while the business market will get the Pro 6000p series. The specs released so far differ only in mentioning that the Pro 6000p series will be supported by the remote secure erase feature of Intel’s Active Management Technology. The third variant—for the embedded and Internet of Things market—will only get the two smallest capacities, which gives us a look at how this design will perform with the limited parallelism that results from using IMFT’s high-capacity 384Gb 3D TLC die.

The 600p and 6000p series are a much more mainstream design than Intel’s previous NVMe SSD for the client market. The SSD 750 was a thinly-disguised enterprise drive, with power consumption and physical dimensions that are far too big for the M.2 form factor that has become the preferred choice for client PCIe storage. The SSD 750 was in many ways overkill from the start, and more recent M.2 drives (especially from Samsung) have caught up in peak performance to offer a much better value for typical client usage. The 600p will be going after the client PCIe storage market from the opposite end: as one of the first TLC PCIe SSDs, its performance specifications don’t set any records but it will be a much more value-oriented product than any of the M.2 PCIe SSDs currently on the market. Intel has confirmed that the 600p and 6000p are using a third-party controller. UPDATE: Allyn Malventano at PC Perspective has uncovered a forum post with an uncensored picture of the 600p. The controller has “SMI” in big letters, suggesting that it is a Silicon Motion SM2260 or relative thereof, but with different markings than the samples Silicon Motion has been showing off at conventions. Intel has also used Silicon Motion controllers in drives like the 540s.

In addition to the SSD E 6000p, there is a new series of SATA drives for the embedded market. The SSD E 5420s series consists of a 240GB 2.5″ drive and a 150GB M.2 drive, both with 3D MLC and full power loss protection. The E 5420s is rated for one drive write per day, a substantial improvement over the 0.3 DWPD rating of the E 5410s or the 20GB/day of the E 5400s.

Moving on to the data center products, the SSD DC S3520 is a new mid-range enterprise SATA SSD for read-oriented workloads and the third iteration of the S3500 series. The M.2 form factor has returned as an option after the DC S3510 series was only offered in the 2.5″ form factor. As with the SATA drives for the embedded market, performance has decreased but endurance has been bumped up from 0.3 DWPD to 1 DWPD. The larger per-die capacity of the 3D MLC has caused the smallest capacity option to increase from 80GB to 150GB, but 1.6TB is still the largest option for the 2.5″ form factor.

(UPDATED) Finally, for the enterprise PCIe space we have the SSD DC P3520. In March the DC P3320 was announced as Intel’s first 3D NAND SSD and the P3520 was mentioned but specifications were not provided at that time. Intel has since decided to only produce the P3520 and to price it close to the level of SATA SSDs. The reduced performance relative to the DC P3500 is a consequence of reduced parallelism at the same capacity that results from using the 256Gb 3D MLC rather than 128Gb 20nm MLC, and the size of this performance regression is a bit dismaying. The DC P3520 is clearly based on the same hardware platform as the rest of the PCIe data center drives, with a familiar layout for the PCB and heatsink evident in the add-in card version.

These new SSDs will have a staggered release over the rest of the year. Starting next week the DC P3520 will be shipping, as well as the 128GB, 256GB and 512GB capacities of the SSD 600p and SSD Pro 6000p. The 2.5″ DC S3520 will ship in early September. The rest are planned to be available in Q4.

ADATA Launches the SD700 External SSD: Dust, Water and Shock Resistant (with 3D NAND)

ADATA last week introduced its third SSD featuring 3D NAND memory. The new SD700 is a dust, water and shockproof drive that has up to 1 TB of capacity as well as a weight of only 100 grams. The SSD uses the USB 3.0 interface and is compatible with the majority of modern PCs.

The ADATA SD700 comes in a metal enclosure with rubber inlays/pads to ensure hermetic sealing and shock resistance. The company plans to offer three configurations of the drive with 256 GB, 512 GB and 1 TB capacities, all featuring up to 440 MB/s read speed (as conditioned by the maximum real-world transfer rate of USB 3.0 interface due to overhead incurred by 8b/10b encoding). The claimed transfer rates of the SD700 are the same as those of the Samsung Portable SSD T3 (which also used 3D NAND), but the real-world performance of the novelty is yet to be discovered. From a compatibility point of view, the external drives are also similar: they can work with Microsoft Windows, Google Android and Apple macOS.

The SD700 from ADATA is based on 384 Gb 3D TLC NAND flash chips made by IMFT and while the SSD maker does not reveal specifics, it is highly likely that the drive uses Silicon Motion’s SM2258 controller (just like other 3D NAND-powered products by ADATA) accompanied by a USB-to-SATA bridge.

ADATA designed and tested its SD700 drive to IEC IP68 standard to ensure that it dust-tight and can operate for 60 minutes while submerged in 1.5 meters of water. In addition, the maker also tested its new external SSD to the U.S. Army MIL-STD-810G516.6 shock and drop resistance standard. The SD700 will join ADATA’s family of external SSDs and will be among the first 3D NAND-based external drives that meet the IP68 and the MIL-STD-810G516.6 requirements.

The endurance of 3D NAND, as well as the rugged design, will make the ADATA SD700 a good choice for users that who transfer large amounts of data often and would like to ensure that their information will not be corrupted either as a result of degradation of non-volatile memory or because of a physical damage. For additional piece of mind, ADATA offers a three-year limited warranty with its SD700 drives.

The ADATA SD700 external SSDs will be available in all-black as well as black and yellow color schemes shortly at Amazon and Newegg. The 256 GB version will cost $109.99, whereas the 512 GB version will be priced at $189.99. At least from a pricing standpoint, the new SSDs from ADATA look very competitive because they are roughly two times cheaper than LaCie’s Rugged Thunderbolt + USB 3.0 SSDs of the same capacity.

Mushkin Launches Reactor Armor 3D and Triactor 3D 2TB SATA SSDs: 3D NAND, SM2258

Mushkin at CES demonstrated its new SSDs in 2.5”/7 mm form-factor aimed at mainstream PCs with a SATA interface. The new Reactor Armor 3D and Triactor 3D use 3D NAND flash memory, the same controller from Silicon Motion and offer nearly similar performance. The main difference is that the former use 3D MLC, whereas the latter uses 3D TLC memory.

The NAND flash industry is transitioning to various 3D NAND architectures that enable higher densities, lower per-bit costs and higher endurance compared to planar flash made using very thin process technologies. So far it has not been easy for independent makers of drives to secure a supply of 3D NAND memory because some manufacturers are cutting down the share of produced flash they sell on the open market, whereas 3D NAND from others does not suit SSDs well. In the recent months ADATA was the only independent supplier of drives to offer 3D NAND-based drives, but as we observed at CES, the situation is about to change. Mushkin is another company to announce a lineup of SSDs featuring 3D NAND and targeting different market segments, from entry-level to the high-end. Unlike ADATA, Mushkin is announcing all of its 3D NAND SSDs at once, which implies that the company can get enough chips for different kinds of drives.

Mushkin’s Reactor Armor 3D and Triactor 3D SSDs are based on Silicon Motion’s SM2258 controller, but while the former uses 3D MLC NAND, the latter uses 3D TLC NAND from an undisclosed manufacturer. The SM2258 controller has four NAND flash channels, LDPC ECC technology, a SATA interface, a DRAM buffer support as well as pseudo-SLC (pSLC) caching in order to maximize SSD performance. At present, the SM2258 is virtually the only market-ready third-party SSD controller with that supports 3D NAND (technically speaking, the SM2256 also supports 3D NAND, but drive makers prefer the more advanced controller so to address the higher end of the SSD spectrum), so Mushkin’ s choice is not surprising if the company needs rapid time-to-market (which is also why it does not wait for Phison’s PS5008-E8). What is even more interesting is that Mushkin is considering to add 3D NAND-based drives to the Reactor lineup that uses the SM2246EN controller (this one is qualified for 3D MLC as well). It does not look like the company has made any final decisions, but it is considering such possibility in a bid to continue addressing the entry-level segment with the Reactor lineup.

Mushkin does not disclose the name of its 3D NAND flash supplier, but we have a reason to believe that this is Micron. SanDisk and Toshiba are shipping their 64-layer BiCS NAND inside their removable media products and promise to use this memory for their SSDs. But as of now, 64-layer BiCS chips have not been qualified for SSDs. 3D NAND from SK Hynix is also available for various products, but it has not been qualified for SSDs just yet.

The Reactor Armor 3D SSDs will be available in 240 GB to 1920 GB configurations, whereas the Triactor 3D drives will feature 256 GB to 2 TB capacities. The former family will take advantage of MLC and offer slightly better endurance albeit at a higher price, whereas the latter lineup will be more aggressively priced thanks to cheaper memory. At the same time, it is noteworthy that both product lines include high-capacity (~ 2 TB) drives, an indicator that they target customers who need a lot of non-volatile memory and can pay for that.

As for performance, Mushkin rates sequential read speed of both Reactor Armor 3D and Triactor 3D drives at 565 MB/s, whereas sequential write speed is rated at up to 525 MB/s and 520 MB/s (when pseudo-SLC caching is used) respectively. Random performance of the drives is specified at up to 90,000 read IOPS and up to 85,000 write IOPS.